What matters in AI.

Subscribe

The Fork-and-Flush method removes chat history and increases scores

On 13 long tasks with the same budget, it got a score of 0.78, and runs without branches got 0.47.

Claimed, not confirmed

Researchers made a method for AI agents that work on long tasks. The name of the method is Fork-and-Flush. The agent divides into K parallel branches. Each branch keeps the files of the workspace but does not keep its chat history. On 13 long tasks with the same budget, the method got an average score of 0.78. Runs without branches got 0.47.

Sources

  1. Clearing chat history helps AI agents find better solutionsthe-decoder.com
  2. Clearing chat history helps AI agents find better solutionsthe-decoder.com
AI MATTER · NEWS · AI MATTER · NEWS ·12 OCT2026

Posted

More in Research

All Research news